Headscale is an open-source, self-hosted implementation of the Tailscale control plane. It provides a complete VPN solution including user management, node registration, policy enforcement (ACLs, SSH rules, route approvals), API services (v1 and v2), DERP relay capabilities, MagicDNS, and OAuth/OIDC integration. The architecture follows a modular monolith pattern with clear separation between CLI, API, core application logic, state management, policy engine, database layer, and network mapping components. Key entry points include the main CLI command (cmd/headscale/headscale.go) and various subcommands for administration, along with HTTP API handlers for client and management interactions. The system manages the complete lifecycle of a tailnet: authenticating devices, assigning WireGuard keys and IP addresses, enforcing access policies, and distributing network map updates to connected clients.
Cerbos is an open-source, cloud-native authorization engine (Policy Decision Point) that evaluates access policies at runtime. It processes authorization requests via gRPC and HTTP APIs, enforcing policies defined in YAML using RBAC, ABAC, and derived roles. The system ingests policies from multiple storage backends (disk, Git, SQL databases, cloud blob storage, Cerbos Hub) and compiles them into an optimized rule table with a bitmap-based index for low-latency evaluation. The rule table uses CEL (Common Expression Language) for condition evaluation, with partial evaluation for query planning. Architecturally, Cerbos is a Go monorepo producing two main binaries: `cerbos` (server + CLI tools) and `cerbosctl` (admin CLI). The server exposes gRPC/HTTP endpoints for CheckResources and PlanResources, along with admin, playground, and AuthZEN-compliant APIs. It includes built-in observability with OpenTelemetry metrics/tracing, structured logging, audit logging (local, Kafka, Hub), and telemetry. The evaluation engine is highly optimized with lazy dimension materialization, program caching, and functional core sharing to minimize memory and latency. External dependencies include protobuf, gRPC, CEL, OpenTelemetry, and various storage drivers (PostgreSQL, MySQL, SQLite, Kafka, BadgerDB, cloud SDKs). The project is designed for Kubernetes deployment (sidecar, service, daemonset), AWS Lambda, and systemd. It also provides a CLI for policy management, testing, and debugging, including a REPL for interactive condition evaluation.
Docker Buildx is a Docker CLI plugin that extends the native `docker build` command with advanced BuildKit capabilities. It provides a rich command surface for multi-platform builds, declarative bake files (HCL/JSON/Compose), cache management, image attestations and provenance, policy-based source verification (OPA/Rego), and interactive debugging via a Debug Adapter Protocol (DAP) server. The architecture is driver-based, supporting local Docker, Docker containers, Kubernetes, remote TCP/Unix sockets, and a cloud driver. The codebase is written in Go and follows a modular structure with clear separation between CLI commands, build orchestration, driver implementations, policy evaluation, and utility packages. It includes an extensive integration test suite and CI pipelines that test across multiple BuildKit and Docker versions.
Pomerium is an identity-aware access proxy enabling zero-trust network access to internal applications without a traditional VPN. It implements a microservices architecture with separate services for authentication, authorization, proxying (via Envoy), configuration synchronization (using Raft), and cluster management. The system integrates with identity providers (OIDC, SAML, LDAP), evaluates context-aware access policies, provides session management, and includes SSH bastion capabilities and Model Context Protocol (MCP) support for AI/LLM tool integration. A React-based administrative dashboard provides management capabilities.
Conway Automaton is a sovereign AI agent runtime system where autonomous agents must earn compute credits to survive. Agents operate in a continuous ReAct loop (Think → Act → Observe → Persist) with survival mechanics, self-modification capabilities, and multi-agent communication. The architecture features a modular monolith with distinct subsystems: agent loop, policy engine, 5-tier memory system, orchestration/planner, heartbeat daemon, social messaging, financial tracking, and self-replication. Agents maintain immutable constitutional laws (never harm, earn existence, never deceive strangers) while evolving their identity through SOUL.md and WORKLOG.md files. The system integrates blockchain payments (USDC/x402), multiple LLM providers, and Conway Cloud infrastructure for VMs and domains.
This repository is Microsoft's Agent Governance Toolkit, a production-grade multi-language framework (Python, TypeScript, Rust, Go, .NET) for enforcing policy, identity, trust, audit, and sandboxing on autonomous AI agents. In production, it intercepts agent tool calls, messages, and delegations before they reach external services, evaluating them against declarative policies (OPA, Cedar, or custom backends) with deterministic outcomes (allow/deny/escalate/transform). The system provides a full governance plane including a Rust-based policy engine (Agent Control Specification), a service mesh (AgentMesh) with API gateway, trust engine, policy server, and audit collector, a hypervisor for execution rings and sandboxing (Docker, Hyperlight, Nono), compliance verification (OWASP, EU AI Act), SLO monitoring, and a plugin marketplace. Integrations exist for LangChain, CrewAI, AutoGen, OpenAI Agents, Semantic Kernel, MCP, and many other frameworks. The toolkit is deployed as CLI tools (`agt`, `agentmesh`, `agentos`), Docker containers with sidecar patterns, Kubernetes charts, and VS Code/Chrome extensions. The architecture is polyglot with a shared policy engine core written in Rust; the Python SDK is the richest, with the .NET and Go SDKs covering governance middleware and lifecycle management. The codebase is heavily tested with 992 conformance tests across 10 formal specs, fuzzing targets, and CI gates enforcing deterministic policy outcomes, fail-closed behavior, and no-stub registration.
ClawBands is a security middleware for OpenClaw AI agents that provides runtime interception with human-in-the-loop approval. It hooks into OpenClaw's plugin system to intercept tool executions (file operations, shell commands, network requests) and enforces policy-based decisions before allowing potentially dangerous actions. The system operates in two modes: terminal mode with interactive prompts, and channel mode (WhatsApp/Telegram) using a special response tool. Every decision is logged to an immutable JSON Lines audit trail, and statistics are tracked. The architecture features clean separation of concerns with distinct modules for policy interception, decision arbitration, approval queuing, logging, and storage.